Bad design examples

[i4004]

here i'll be putting examples of psus that failed (either catastrophically or otherwise). at least one supply from here fried rest of the machine.
(worst cases will be mentioned last, not because i want to keep you waiting/as i'll post this over few days/ but because i took pix in that order and made folders in that order..and..i need some order... ).

by doing this perhaps we can learn what to avoid when it comes to purchasing new psus.

so let's start. somewhat typicall example, this unit doesn't have EMI filter at all.
primary rectifier is comprised of four diodes, uses cheapest caps available(as i mentioned elsewhere, teapos look like premium grade compared to this junk..all of these psus have bad caps even if pix don't represent it fully), pwm ic is sg6105.

i don't have housings for most of these (they take too much space) but these are easy to spot; they're unusually cheap, they're very light, they bare weird brands etc.

[i4004]

Re: Bad design examples

2nd specimen: here we see something that repeats often at these cheap and cheesy designs->temperature strain around the two small transformers (+5V stby trafo and another one that connects low voltage secondary circuit (its pwm section) to primary, to reach and steer those hi-power switchers, but still keep primary and secondary sections electrically separated).

the only explanation i can find for this are (surprise, surprise) badcaps. the more they work the warmer they get and then they lose some more of their crappy electrolyte and this vicious circle continues untill their esr reaches astronomical values and circle just stops functioning, or untill that causes some disaster like (quite literally) burning itself and other components(like we'll see in examples to come) or failing to steer those switchers, which can lead to producing, well, anything on the load, and usually it's nothing good.

this design is simillar to above shown, has some emi filtering, sam pwm chip, and with lytics like this no good chance of reliably lasting.

as we'll see later, these are actualy better examples(lasted some years, didn't kill no mobos, didn't produce smoke). we'll see worse. much, much worse.

[PCBONEZ]

Re: Bad design examples

psst: i4004, ,, It helps if you tell make/model..

[starfury1]

Re: Bad design examples

he probably doesn't want to get done for slander and liable even if it is TRUE
Yep that is a joke

Cheers

[i4004]

Re: Bad design examples

like i said, i don't have cases, as they take space. i was just preserving pcbs.

these are psus of this
http://www.jonnyguru.com/modules.php...=Story&reid=71
sort.

i know mfrs of 3 of them, though, and will surely mention them.

but after all, they're all no-name...hehe...
model, you say? it's probably all one model with different stickers on it.
<wink>

[starfury1]

Re: Bad design examples

yeah all provided by the Wun hung low company and stripped to to the bear min to save a few cents. thanks i4004 for posting and bringing the bad points together in one thread

Cheers

[i4004]

Re: Bad design examples

i'm feeling generous, so let's do 2 today. next time we'll finish it off.

3rd one shows more prominent secondary caps deterioration. again, it's better to have secondary(output) caps bulging and pwm shutting down ocassionally(making you swap the psu while it's still on time) than to have crap happening on primary that causes your mobo to fry. this psu didn't cause catastrophic failure that ate whole machine. it deteriorated gracefully. it's crap, but still, worse can happen.

this design employs different pwm ic, but i wouldn't say failures are connected to ics at all. it's more about passive components, their quality, and the lack of them (emi filters missing etc.). these stress out switchers, and they fail. upon failing they can produce too high voltages on output. nothing can prevent disaster, as blown transistor cannot be regulated.
(that is, at least, best explanation i came with thus far. if anybody has better explanation i'm interested to hear it.)
at first i was thinking it may be a flaw of transformer, but that is probably least probable, even though i have actually seen one example of trafo with primary shorted to secondary, but even experienced repairer will be surprised by such outcome. it is completely unusual event.

[i4004]

Re: Bad design examples

moving on to case no.4: here it gets interesting, as we'll see marks of real burning on this one. it seems here we had a bad cap on 5v stb primary section (T703 place) that caused smoke coming from psu, and this made owner swap it.
nothing bad happened to the rest of the machine.
cap manufacturer is "E.V.A.top", it's shown on this list
http://capacitor.web.fc2.com/abcde.html#abcde
no website, it seems, i surely hope they died. <wink>

you can also see one more evatop cap on the secondary side, not completely destroyed as one on the primary side, but bulging.

this psu was called "sunny", if i remember correctly.
http://www.sunny-group.com/en/production.asp
i dunno if they're making such psus anymore, it seem they don't, good for them and us!

next time we'll see same design(same pcb actually) with another brand, but equally crappy, the one that destroyed my complete system back in 2001.
we'll devote some more time to that...

[i4004]

Re: Bad design examples

something else belongs above (case no.4) : esr measured on those caps.
on pcb you have numbers, and those numbers correspond to particular esr measurement written on paper.
those evatops are crap.

[davmax]

Re: Bad design examples

this design employs different pwm ic, but i wouldn't say failures are connected to ics at all. it's more about passive components, their quality, and the lack of them (emi filters missing etc.). these stress out switchers, and they fail. upon failing they can produce too high voltages on output. nothing can prevent disaster, as blown transistor cannot be regulated.

I am not sure what you are saying here.

1. EMI filters are to simply stop switching RFI getting on to AC power lines, they have no real affect on the switching transistor that is fed by the main high voltage DC filter capacitor. The EMI filters are before the AC rectification that supplies the filter cap.
2. When the switching transistor blows there is no output via the transformer and therefore no excessive voltage.
3. Excessive voltage can only occur if switching continues and there is a failure in the feedback control loop that will cause excessive duty cycle on each switching period. Could be a reference voltage gone wrong, bad connection, failed resistor, one of number of possibilities. A faulty cap on one output pulling down that voltage can raise the voltage of other outputs in that most computer SMPS operate on the sum of several outputs.

[i4004]

Re: Bad design examples

1-emi filter also can muffle and dampen voltage spikes that come via the mains, and it can limit the current inrush upon hitting the 'on' switch.
if the circuit is already working out of spec (because its caps are bad) such things can be decisive thing that bring the card-house down. looking at the filtering section of fortron psu(https://www.badcaps.net/forum/attach...achmentid=7338) makes me much more convinced that on-event won't break silicon than looking at these that have no emi filter at all...and also better than enermax emi filter (https://www.badcaps.net/forum/attach...achmentid=7268), for that matter. circuit is called emi filter, but it does more than that. here's an excerpt from the "Practical Switching Power Supply Design" book

3.1 The EM1 Filter
This section is composed of a small L-C filter between the input line
and the regulator. It serves a dual purpose. First, C, and L, act as a
high frequency radio-frequency interference (RFI) filter, which reduces
the conducted high frequency noise components leaving the switching
suppty back into the input line. These noise currents would then radiate
from the input power lines as in an antenna. The lowpass cutoff frequency of this filter should be no higher than 2 to 3 times the supply's
operating frequency. The second purpose of this stage is to add a small
impedance (L,) between the input line and the bulk input capacitor. It
basically reduces any lethal transient voltage and allows the bulk input
filter capacitor and any surge protector to absorb the destructive energies from the input line spikes or surges with little chance of exceeding any of the components' voltage ratings.

2-when the event is over, sure there's nothing on the output, but what about when you have that initial current inrush that breaks the switcher? what does that cause in the trafo core and how does it affect secondary? one big current spike is enough to fry anything on the secondary.

3-excessive output voltage can also occur in the event of nothing stopping the primary current from going over the top.

[davmax]

Re: Bad design examples

1-emi filter also can muffle and dampen voltage spikes that come via the mains, and it can limit the current inrush upon hitting the 'on' switch.
if the circuit is already working out of spec (because its caps are bad) such things can be decisive thing that bring the card-house down. looking at the filtering section of fortron psu(https://www.badcaps.net/forum/attach...achmentid=7338) makes me much more convinced that on-event won't break silicon than looking at these that have no emi filter at all...and also better than enermax emi filter (https://www.badcaps.net/forum/attach...achmentid=7268), for that matter. circuit is called emi filter, but it does more than that. here's an excerpt from the "Practical Switching Power Supply Design" book


2-when the event is over, sure there's nothing on the output, but what about when you have that initial current inrush that breaks the switcher? what does that cause in the trafo core and how does it affect secondary? one big current spike is enough to fry anything on the secondary.

3-excessive output voltage can also occur in the event of nothing stopping the primary current from going over the top.

1. I agree that the EMI filter also acts to supress input power transients. However this is probably the highest area of reliability in a PSU and the transient has to be particularly massive to raise the input filter cap to any great magnitude bearing in mind the limiting action of the filter. The failure of EMI filtering caps has little effect on the supression of input surges, these surges/in rush are limited by the inductance and a transient suppressor both protecting the input filter cap. In simple terms this is a low risk area. In the extreme voltage case and the transient supressor failing then action in 2. may occur.

2. This frying the secondary with one big transient is speculation. I do not believe it. If the MOSFET goes short the excessive current flow will saturate the transformer, limiting the energy available and at the same time blowing the input fuse. There will not be enough energy to raise all outputs significantly for any substantial duration to do damage. Are there any MOSFET shorted failures proven to have caused damage due excess output voltage? This is not something I am aware of.
3. Over voltage problems are more likely due to sustained excess.

You are doing a great job presenting all the PSU photos and supporting info about their failures.

[i4004]

Re: Bad design examples

dav, here's the case no.5, the one which should put your doubts to rest.

first we'll see pix of it, and in next post (autopsy) we'll try to figure out what happened to it.

let's compare this(https://www.badcaps.net/forum/attach...achmentid=7351) and the previous(https://www.badcaps.net/forum/attach...achmentid=7323). obviously, same pcb, but the case no.5 has few components missing.
(i took out c311, but the rest is as-is. we can see missing transistor, missing resistors that accompyny it, 2 missing caps...).

notice green resistor that is completely black now(burnt).

[i4004]

autopsy of case no.5

now we arrive at interesting part. trying to establish what happened here. first we'll see the results and then we'll see if we can figure it out.

these are parts from the +5v stb circuit that provides psu and the mobo with initial power to operate. it's oscillator that's hooked to this small trafo(T703 shown in case 4), and output of it is rectified and supplies pwm ic and mobo.
this design is simple (really too simple) and it obviosuly offers no safety.
notice dm311 on my new fortron(https://www.badcaps.net/forum/attach...achmentid=7346), it's this chip
http://www.fairchildsemi.com/pf/FS/FSDM311A.html.

moving on to pix that show the following:

-jenpo 4.7uF lytic with substenitaly increased esr (about 25ohm)
-blown switching transistor (2sc5353)
-even the transformer is blown (it's first pin and attached wire chared and cut to 2)
-chared resistor (it's purpose was probably to provide a way for excess current to escape, but it didn't do this job too well...)
-mains fuse (not shown here) is intact. obviously it has thicker wire than this trafo.

what seems to have happened: crappy jenpo cap is gradually increasing its esr. when it reaches a certain level it causes oscillator to stop functioning, and this means there's nothing to properly limit the current that flows thru the transistor and trafo. during a short period transistor is not switching at all(it's blown), this increses the primary current to max the trafo wire can take and in the end that wire fails. offcourse this doesn't go unnoticed on the secondary side: mobo is destroyed alongside everything that's connected to it.
(main switchers are ok, so it can be discussed if this 5v stb failure also induced pwm ic to go out of spec and produce crap on main trafo primary and on all rails, but that interest me less, as i'm interested in how things unfolded. and i'm pretty sure 5v stb circuit was the trigger...it's obviously very bad design if we see it failing on 2 psus, cases 4 and 5).

that is my version anyway. others can offer better explanations if they have 'em.

this was jnc psu.
http://www.jnc.com.tw/company.htm
---------------------

now the comments regarding davmax:

-this circuit doesn't have feedback. at all.
so it wasn't the control loop.

-i don't think sustained excess would fry trafo windings. i also don't think trafo would fail on its own(like i explained earlier). sustained excess would be good explanation for frying of example 4, where it's obvious temperature strain was prolonged. but transistor and trafo never failed. here we see effects of sudden blow. not much of slow burning, but a lot of damage.

-as for emi-filter, yeah, i agree i probably put more accent on it than it needs it. filter cap is the one taking most of the beating if emi filteris missing, and these caps usually don't fail, so i'll agree it's "low risk area" as you've put it.
the fact is that this happened soon after turning the machine on, but it didn't happen at that instance, but a bit later. i know some time has passed between turning it on and its demise. i turned it on and left the room. when i returned it was dead.

i'll now add one more psu and then few conclusions.

[i4004]

Re: Bad design examples

moving on to our last patient, propower psu from 1993. no.6.

built around dbl494( same chip used for a whole decade, it seems. cases no.4 and 5 are early 00s), but a better design, not so much saving is apparent. again a lot of the badcaps. this supply failed but didn't destroy machine, and as such it's not too interesting subject to investigate.

looking at these pix i notice 7805 regulator and it reminded me to search for it on case no. 5 too: and ineed it has one, so i should be clear:
7805 is used to supply mobo with stby voltage so it will not be the one that destroyed the mobo in the case no.5. but the pwm ic indeed is supplied straight from that trafo, after being rectified and smoothed. ic accepts supply voltages of 7-40V.
that doesn't change the main reason of failure, but it means that more than likely this is correct:
this 5v stb failure also induced pwm ic to go out of spec and produce crap on main trafo primary and on all rails
from above post.

i still think same thing (one big blow) can produce too high voltage on the secondary of main trafo, in a same way it did for this small trafo.
i should contact johnnyguru or torres to make them record event next time they blow cheap psu.

anyway, the pix.

[i4004]

Re: Bad design examples

aside from usual conclusions(too cheap, too light psus usually are crap that can be potentially dangerous) interesting thing is that i would say putting good caps here would prevent most of these failures.
not all, and this wouldn't make these psus produce more power, or anything like that, but anytime you open them up you see a bunch of badcaps.
thier design is such that they're loading these caps too much, but good caps probably wouldn't fail on same spots.

browsing thru the newegg consumer reviews, and comparing experiences of cheap vs. better quality psus it's visible that even cheapest crap rarely fails in a drastic ways.
which means overall we have small number of psus damaging the machines, and because of that i would encourage anybody who has such psu to dismantle it, and show us what's inside.

[davmax]

Re: Bad design examples

i4004. There are some fundamentals missing in your inputs.

1. ALL SMPS have a feedback loop to ensure control of the output voltage. Refer to FSDM311A data sheet page 2 where feedback loop connects to pin Vfb.
2. SMPS operating at 50 - 100Khz use small low energy transformers delivering energy pulses every 10-20uS ie many relatively small energy pulses/sec achieving high power unlike a 60Hz tranformer that delivers fewer higher energy pulses(half sine) and the transformer has to be correspondingly larger. Now when heavy primary current flows the transformer core will saturate at max flux density, this respresents the max energy a transformer can store and therefore transfer to the secondary. Now considering a SMPS transformer that is low energy, when it saturates it may store up to 50% more energy for that one pulse event compared with the max of one 10-20uS pulse. So in the scheme of things this is a very small addition to the total energy/sec and will not supply significant extra energy to raise output cap voltages by any large amount. So what I am saying is that heavy current in a transformer primary only transfers a small proportion of energy across the transformer. Normally a primary failure will blow the input fuse, if no fuse then something else will blow.
If there is a blow out in the secondary side it is more likely due to an insulation failure in the transformer typically caused by the primary winding overheating.

[davmax]

Re: Bad design examples

this 5v stb failure also induced pwm ic to go out of spec and produce crap on main trafo primary and on all rails

Is there any evidence of this? Normally the 5V stb is independent of the main switch mode regulator ie not involved in the control of operation other than switching it ON.

[i4004]

Re: Bad design examples

1. ALL SMPS have a feedback loop to ensure control of the output voltage. Refer to FSDM311A data sheet page 2 where feedback loop connects to pin Vfb.

i mentioned my new fortron psu and its fsdm311 chip as an example of better design than one i was describing.
i said the "circuit" didn't have feedback, ie circuit that is providing initial power to pwm ic, which in the case of crap psu is db494.

look at the left-bottom of the image: "second power supply".
and..huh...i should of looked at it earlier to, as it seems it has some sort of feedback via other section of primary windings.
d29 and elko c19 probably rectify and smooth this error voltage and lead it to base of q12.
(this schematic doesn't show same thing as this psu hosts, but it's fairly simillar)

also, it was a mistake to call this circuit "+5v stb circuit", because it's really not it. i mistakenly thought this circuit is supplying both pwm ic and +5V stb to the baord, but it's not. it's just supplying the ic.
+5V stb has another path alltogether, and as i said, it's protected by 7805 regulator.

So what I am saying is that heavy current in a transformer primary only transfers a small proportion of energy across the transformer.

so what explanation on why "second power supply" failed you offer?
by your account, high primary current can't wreak havoc on secondary side.
so what did it?
to repeat again, this transformer is not the main smps trafo(its ok, and main switchers are ok), but the auxillary one, one that supplies power to pwm ic because first you need to start ic.

we only have proof of high primary current, because transistor and trafo(primary section) are blown. this means we surely had high current there.
my suggestion is that this caused voltage spike that offset pwm ic and then it sent wrong signals to main switchers.

the other path we could take would be to discard what was happening to this secondary power supply alltogether, but seeing how it also caused problems on psu no.4 i think we shouldn't. supply to pwm ic is important.

btw. what's the source for saturation being 50% more energy?
do you have graph representing flux.density vs. current?

[davmax]

Re: Bad design examples

i4004. The circuit you have supplied is incomplete in many connections making it difficult to for example see where the supply voltage to pin 12 of the TL494 comes from, it is connected to nothing!!! Clearly it must be connected to a supply independent of the main switcher so that it can start and drive the switcher.

What the circuit diagram does show is that the second power supply is entirely seperate from the main switcher and obtains it's power from the primary DC input rails. It also shows that it is most likely a free running oscillator without any feedback regulation because it only has to supply the 7805 for a 5Vstb. There is no relationship as I pointed out to the secondary or primary circuits of the main switcher and cannot therefore screw up the main switcher as you stated. Or the other way round, the switcher screw the secondary power supply.

You ask why would this secondary supply fail? Looking at the circuit the most likely failure would be the high voltage transistor switching this small transformer.

I hope my explanation of the blocks of energy being switched was understood. This taken from electronics engineering fact. There is no way a massive main switcher primary current can all be transferred across the transformer. If you do not understand my explanation please study transformer theory and operation or get somebody who knows to explain it.

btw. what's the source for saturation being 50% more energy?
do you have graph representing flux.density vs. current?

The extra energy comes from the difference between the transformer core flux density at the normal maximum current and the current value at saturation. If for example it required 30% more primary current to saturate the core the increased and maximum stored energy available increases by 69% to be precise. Because the energy increases at the rate of I squared.
The principle I need you to grasp is the fact that a low energy transformer transfers energy in small energy blocks at the clock frequency (say 50-100Khz) so if one block is even twice it's normal energy it is still not going to have a major impact. A switching transistor short circuit will produce only one pulse energy block to saturation level, any current beyond that point will not increase the core flux density and therefore stored energy, inductive reactance to current flow will cease and the short circuit current will rapidly increase blowing anything in it's path.
I am not going to give you a long explanation of transformer theory, if you do not know you should be very careful not to mislead yourself and others.